Research On Privacy Preserving In Social Networking Based On Graph Modification And Clustering

While enjoying the convenience and entertainment of social networking, individual are suffering from privacy disclosure at the same time. This thesis combines with undertaken major science and technology project to research the related technique of privacy preserving, which has important researching significance and application value.This thesis first proposes the privacy preserving technique based on maximal frequent sub-graph miming and graph modification. This technique first uses Margin mining algorithm to find the maximal frequent sub-graphs. Similar sub-graphs form a group and modification on these sub-graphs properly brings out K-isomorphism anonymous graph, which preserves privacy when published. The data through such process is applied to study the local structure characteristic of social networking. The experiments show that the algorithm retains much information of the local structure of the social network and achieved a good balance between privacy security and data utilize.Then, this thesis researches the privacy preserving technique based on signaling model and clustering. The algorithm first calculates the similarity and distance between individual in the network based on the signing model. Then nodes are divided into groups by clustering algorithm. At last each cluster is generated into super node, forming the K-cluster anonymous graph which preserves privacy when published. The data processed by this algorithm is applied to study macro properties of social networking. The experiments show that the data processed by this algorithm has a better utility comparing with others while protecting individual’s privacy effectively.Finally, the privacy preserving algorithm oriented the local structure and the algorithm oriented the macroscopic properties are applied to the Qiantang communications platform. Then the design of the data publish system in the platform is introduced in detail. The two different application-oriented algorithms form the data publishing system. In the end the thesis shows the effect diagram of the application of the two algorithms.